Methane emissions decreased in fossil fuel exploitation and sustainably increased in microbial source sectors during 1990–2020

Rapid methane mitigation could avoid approximately 0.3 °C of warming by the 2040s, and over 100 countries have pledged a 30% reduction in anthropogenic CH₄ by 2030. Effective mitigation requires robust attribution of sectoral and regional sources. Traditional bottom-up inventories and process models provide sector resolution but are uncertain in emission factors and activity data, while top-down inversions match atmospheric observations but struggle to disentangle source sectors. Stable isotope ratios of methane (δ¹³C and δD) add source attribution information because microbial and thermogenic sources have distinct signatures. However, prior isotope-constrained studies have yielded divergent conclusions on the drivers of changes in atmospheric CH₄ since the 1990s. This study aims to reconcile atmospheric CH₄ trends and isotopic observations by testing sector-specific emission scenarios within a 3D chemistry-transport framework and by quantifying uncertainties from source signatures and kinetic isotope effects.

Previous 3D inversions and box-model studies that include δ¹³C-CH₄ have suggested either dominant increases from microbial sources or combined increases from microbial and fossil fuel sources to explain recent methane growth. Bottom-up inventories (e.g., EDGARv6, GAINSv4) suggest rising fossil fuel emissions due to unconventional gas, coal mining, and economic growth in Asia, but these trends often conflict with atmospheric constraints and regional inversions. Discrepancies arise from uncertainties in emission factors (especially fugitive fossil fuel), process-model inputs for natural sources (e.g., wetland area and biogeochemical parameters), sparse δ¹³C observations, uncertain δ¹³C source signatures by region/sector, kinetic isotope effects of sinks, and sampling biases. Modeling choices in inversions (optimizing emissions vs. signatures) also lead to different attributions. Thus, a comprehensive forward-model assessment testing multiple emission inventories, isotope parameters, and sink representations is needed to evaluate consistency with observed CH₄ and δ¹³C trends, gradients, and profiles.

The study uses the MIROC4-ACTM to simulate CH₄, δ¹³C-CH₄, and δD-CH₄ globally from 1970–2020 at ~2.8° resolution with 67 vertical levels. Meteorology is nudged to JMA reanalysis. Four emission scenarios are constructed: (E₀) baseline using EDGARv6 for anthropogenic sectors, VISIT for wetlands/rice, EDGARv6 for ENF&MNM and landfills, GFEDv4s/MacCity for biomass burning, and Etiope et al. for geological CH₄ (fixed annually); (E₁) replaces EDGAR ONG with GAINSv4 ONG; (E₂) further removes U.S. unconventional gas emissions after 2006 from GAINSv4 ONG and scales China coal emissions based on regional inversion trends; (E₃) reduces global geological emissions from 37 to 19 Tg CH₄ yr⁻¹ while keeping E₂ fossil adjustments. Tropospheric OH fields (scaled Spivakovsky et al.) are used with no long-term OH trend; Cl fields are tested using a control (Cl_ctrl) and an alternative (Cl_wang) with enhanced tropospheric reactive chlorine; O('D) is computed online. Kinetic isotope effects (KIE) for OH are tested using low (1.0039) and high (1.0054) values; the effect on δ¹³C magnitude and gradients is assessed. Initialization derives 1970 3D fields from observation-based states and Rayleigh fractionation, with spin-up (CH₄: 15 years; δ¹³C: 25 years). Model evaluation uses surface CH₄ and δ¹³C-CH₄ from NOAA/INSTAAR and TU/NIPR at marine boundary layer sites aggregated into three latitude bands; δ¹³C datasets are harmonized with an offset. Vertical profiles from balloon measurements over Kiruna, Sweden and Hyderabad, India are used to test vertical transport and stratospheric chemistry. Trends and growth rates are extracted via harmonic fitting and low-pass filtering. Additional sensitivity ensembles (32 runs) vary sector-specific δ¹³C source signatures using geographically varying maps (coal, ONG by country; ruminants and biomass burning by C3/C4 fractions; wetlands by subtype and C3/C4 pathways) versus globally invariant values, combined with the two KIEOH choices.

- The baseline scenario E₀ (EDGARv6 increasing fossil and microbial emissions) matches CH₄ growth before 2000 but overestimates post-2000 growth and produces a δ¹³C-CH₄ trend opposite to observations (model increases, observations decrease from mid-2000s), indicating inventory inconsistencies. - Adjustments targeting fossil fuel sectors best reconcile CH₄ and δ¹³C simultaneously. Replacing ONG trends with GAINSv4 (E₁) still overestimates post-2000 growth and mismatches δ¹³C trends, implicating excessive U.S. unconventional gas and/or China coal emissions in inventories. - Excluding U.S. unconventional gas increases since 2006 and scaling China coal growth to regional inversion trends (E₂) yields high correlations with observed CH₄ (R > 0.85) and δ¹³C (R > 0.7) trends, but with a residual CH₄ offset. - Reducing global geological emissions from 37 to 19 Tg CH₄ yr⁻¹ (E₃) removes the offset and reproduces observed CH₄ and δ¹³C long-term trends and magnitudes across latitudinal bands. - Decadal fossil fuel (FF) emissions decreased by about 6 Tg CH₄ yr⁻¹ from the 1990s (129 ± 7 Tg yr⁻¹) to the 2000s (123 ± 4 Tg yr⁻¹), driven by reductions in ONG emissions associated with decreased oil production and improved practices (e.g., reduced venting and enhanced recovery) in regions such as Russia and parts of Africa. - From the 2000s to 2010s, FF (~124 Tg yr⁻¹) and biomass burning (~33 Tg yr⁻¹) remained relatively stable, while microbial emissions increased by ~27 Tg CH₄ yr⁻¹, largely from enteric fermentation/manure and waste in developing regions, with wetlands contributing ~16% of the microbial increase. - Spatial δ¹³C source signature maps (wetlands, ruminants, biomass burning, coal, ONG) substantially improve modeled δ¹³C seasonal cycles, interannual variability, and latitudinal gradients relative to globally invariant signatures (r > 0.85 in NHL), without requiring emission changes. - The choice of KIEOH strongly affects δ¹³C magnitude and gradients but has negligible effect on long-term δ¹³C trends; higher KIEOH values are consistent with observed vertical and N–S gradients. - Vertical profiles of CH₄, δ¹³C-CH₄, and δD-CH₄ are well reproduced (r > 0.9), confirming realistic transport and stratospheric chemistry; vertical gradients are sensitive to Cl fields. - E₃ emissions align with independent inversion-based total CH₄ emissions; sectoral estimates for 2018–2020 agree with external constraints (e.g., global coal ~30–32 Tg yr⁻¹; U.S. ONG ~7–8 Tg yr⁻¹ after removing unconventional increments).

By jointly matching CH₄ and δ¹³C-CH₄ trends, latitudinal gradients, and vertical distributions, the study demonstrates that sustained increases in microbial emissions, not rising fossil fuel emissions, primarily drove the post-2006 atmospheric CH₄ growth. Fossil fuel emissions decreased from the 1990s to 2000s and then remained approximately stable through the 2010s, as declines in ONG emissions offset coal increases in China. Incorporating geographically varying δ¹³C source signatures obviates the need for ad hoc emission adjustments often invoked in inversion studies to match spatial isotope gradients, highlighting the importance of accurate isotope parameterization. Sensitivities to KIEOH and tropospheric Cl primarily influence δ¹³C magnitude and gradients, underscoring sink-related uncertainties that do not, however, change the sign or timing of the long-term δ¹³C trend. The findings reconcile conflicts between inventories and atmospheric observations: EDGARv6’s steadily increasing ONG is inconsistent with observed CH₄/δ¹³C, while GAINSv4 trends, after removing debated U.S. unconventional gas increases and moderating China coal growth, are consistent with multiple observational constraints. This sector-resolved attribution supports prioritizing mitigation in microbial sectors (agriculture and waste) alongside continued improvements in fossil fuel management.

Forward simulations using MIROC4-ACTM, tested against decadal, latitudinal, and vertical observations of CH₄ and δ¹³C-CH₄, indicate that fossil fuel emissions decreased from the 1990s to the 2000s and then stabilized, while microbial emissions increased steadily, driving the renewed CH₄ growth since the mid-2000s. The most plausible scenario involves lower ONG emissions than EDGARv6, exclusion of large U.S. unconventional gas increases, moderated growth in China’s coal emissions, and reduced geological emissions (19 Tg yr⁻¹). Incorporating spatially varying δ¹³C source signatures substantially improves δ¹³C gradients and variability, and higher KIEOH values are compatible with observed constraints. These results clarify sectoral drivers critical for methane mitigation strategies: sustained attention to agricultural and waste emissions is warranted, while maintaining and enhancing best practices in oil and gas can preserve the observed stabilization of fossil-related emissions. Future work should refine δ¹³C source signature maps and magnitudes, better constrain KIE and tropospheric Cl spatial-temporal variability, expand isotope observation networks, and integrate alternative wetland models and OH variability to further narrow uncertainties.

Key limitations include: reliance on forward modeling rather than full multi-parameter inversions; uncertainties in sectoral inventories (especially fugitive fossil fuels and wetlands) and δ¹³C source signatures by region/sector; sparse and non-uniform δ¹³C observational coverage and interlaboratory offsets; uncertain kinetic isotope effects for OH and poorly constrained tropospheric reactive chlorine fields; assumed absence of long-term OH trends; and limited ability to distinguish conventional versus unconventional ONG emissions due to shared isotope signatures. These uncertainties can affect absolute source attributions and magnitudes, though the inferred temporal trends (stable fossil, increasing microbial) are robust to tested sensitivities.